| { | |
| "model_name": "fireredpunc", | |
| "framework": "pytorch", | |
| "target_hardware": "AX650", | |
| "npu_mode": "NPU1", | |
| "compiler": "Pulsar2 7.0 via Magnetar COMPILE", | |
| "quant": { | |
| "activation": "U16", | |
| "weight": "S8", | |
| "gather_embedding": "FP32", | |
| "smooth_quant": true | |
| }, | |
| "seq_len": 128, | |
| "inputs": [ | |
| { | |
| "name": "input_ids", | |
| "shape": [1, 128], | |
| "dtype": "int32", | |
| "layout": "NCHW" | |
| }, | |
| { | |
| "name": "attention_mask", | |
| "shape": [1, 128], | |
| "dtype": "float32", | |
| "layout": "NCHW" | |
| } | |
| ], | |
| "outputs": [ | |
| { | |
| "name": "logits", | |
| "shape": [1, 127, 5], | |
| "dtype": "float32" | |
| } | |
| ], | |
| "layout": "[CLS] + tokens, no [SEP]; logits row i belongs to token i", | |
| "labels": [" ", "，", "。", "？", "！"], | |
| "files": { | |
| "axmodel": "punc.axmodel", | |
| "tokenizer_dir": "tokenizer/", | |
| "vocab": "vocab.txt", | |
| "out_dict": "out_dict" | |
| }, | |
| "source": "FireRedTeam/FireRedPunc (Apache-2.0), chinese-lert-base", | |
| "notes": "Input must be unpunctuated; strip ，。？！、,.?! before inference." | |
| } | |